What This Chart Shows
This chart shows you if your next dollar (or next unit of effort) is weaker than what you’re getting on average today. For each lever or group it shows two bars: one for current average return (from 0x to 1x spend/activity) and one for next marginal return (the extra return as you go from 1x to 1.5x).
If the marginal bar is above the average bar, the lever has room to scale. If it's below, the lever is saturated (diminishing returns). When they're about equal, you're in the efficient range. An icon on each row (● efficient, ↑ scale, ⚠ saturated) summarizes this so you can quickly see which channels can handle more budget and which can't.
Key Questions This Chart Helps Answer
- Which marketing channels can handle more budget, and which are already saturated?
- Is the next dollar (or next effort unit) weaker or stronger than my current average return?
- Where should I scale spend vs hold vs optimize?
Axes, Metrics, and Units
Element | Description |
|---|---|
| X-axis | Lever or group name (e.g., Series, Channel, Funnel Stage), depending on grouping mode. |
| Y-axis | Costs mode: ROI per multiplier unit No-costs mode: Lift per multiplier unit Negative returns are clipped to 0. |
| Y-axis (secondary, optional) | Break-even (days): payback period in days when costs are present; diamond markers and “Break-even (days)” label. |
| Bars | First bar (blue): Current average return (0x to 1x). Second bar (e.g. orange): Next marginal return (1x to 1.5x). |
| Status icon | ● Efficient range (marginal ~= current, ratio between 0.95 and 1.05). ↑ Scale opportunity (marginal > current, ratio > 1.05). ⚠ Saturated (marginal < current, ratio < 0.95). |
| Break-even line | Costs mode: Dashed horizontal line at ROI = 1.0. No-costs mode: Zero baseline. |
Control Options Reference
Control | Meaning |
|---|---|
| KPI classification / hierarchy / drilldown | When a KPI is selected, the chart uses that target’s response-curve and attribution block. |
| Grouping mode | Lever, Series, Channel, Funnel Stage, Source Type, or Topic. One row per lever or per aggregated group. |
| Event categories | Filter by specific event categories (include/exclude). |
| Funnel stages | Filter by specific funnel stages (include/exclude). |
| Search terms | Text search on lever/series labels. |
How to Interpret the Results
- Marginal bar above average bar = Scale opportunity: next unit of spend/effort returns more than the current average; consider increasing budget.
- Marginal bar below average bar = Saturated: next unit returns less than the current average; diminishing returns; consider holding or reallocating.
- Marginal bar close to average bar = Efficient range: roughly constant returns; maintain or optimize rather than big scale-up or cut.
- Icons: Use ● / ↑ / ⚠ for a quick read; they reflect the same ratio rules (0.95 to 1.05 = efficient, above 1.05 = scale, below 0.95 = saturated).
- Break-even (costs mode): Bars above the 1.0 line are above break-even; use with scale/saturated to decide scale vs optimize vs cut.
- Break-even (days): When shown, lower days = faster payback; “Never” or missing when the lever doesn’t break even. Use to compare speed of payback across levers.
- Capacity view: Only non-negative returns are shown; the chart is about "who can absorb more budget," not signed cannibalization (see other charts for that).
Practical Applications for Marketers
Application | How to use this chart |
|---|---|
| Budget allocation | Shift spend from ⚠ Saturated rows toward ↑ Scale opportunity rows. |
| Scale decisions | Use ↑ and bar height to choose where to test or add budget. |
| Optimization focus | Use ● and ⚠ to decide where to improve efficiency (targeting, creative) before adding spend. |
| Payback comparison | When break-even days are shown, compare across levers for cash flow and risk. |
| Portfolio view | Group by Channel or Series to see saturation mix by channel or segment. |
| Planning | Use status and bars to set expectations: scale where ↑, hold or optimize where ● or ⚠. |
Common Mistakes and Misinterpretations
Mistake | Why it is a problem | How to avoid |
|---|---|---|
| Treating “Saturated” as “bad” in all cases | A saturated lever can still be profitable and large; it just has less headroom. | Use status for direction (scale vs hold vs cut), not as a single good/bad label. |
| Ignoring the break-even line | Levers below ROI = 1 lose money on average; scaling them can worsen ROI. | Always consider break-even and bar level together when deciding scale. |
| Comparing current vs marginal across different grouping modes | Lever-level vs channel-level ratios are different; group aggregation uses ratio-of-sums logic. | Compare within the same grouping and same filters. |
| Assuming all levers are shown | Only top N rows by impact (max of current, marginal) are shown. | Use filters or grouping to focus; if a lever is missing, it may be low-signal or filtered out. |
| Reading negative returns on this chart | This view clips to a “capacity” view (≥ 0); negative attribution is not shown here. | For negative effects use signed attribution or waterfall charts. |
Caveats and Considerations
- Data dependency: If attribution is missing, the chart shows an error.
- Outcome / KPI block: When an outcome group or KPI is selected, the chart uses the matching block of attribution and response curves. If that block is unavailable, an error message is shown.
- Thresholds: Efficient range uses ratio between 0.95 and 1.05; scale > 1.05; saturated < 0.95. Low-signal rows (below a signal floor) are classified as saturated to avoid noisy ratios.
- Capacity view: Current and marginal are clipped to ≥ 0 for display.
- No-costs mode: When cost data is absent or zero, the chart shows lift (not ROI), break-even at 0, and no break-even days.
- Top N: Only up to 15 rows are shown (by max of current and marginal); use grouping and filters to see the slice you care about.
- Break-even days: Computed from attribution and cost; “Never” or inf when the lever doesn’t break even. In grouped mode, break-even is a cost-weighted average across levers in the group.
- Uncertainty: The chart shows point estimates from the model; it does not show confidence intervals.